Re: [RESEND PATCH v9 6/6] ASoC: amd: Added ACP3x system resume and runtime pm

[vishnu <vravulap@amd.com>]

On 19/11/19 6:26 PM, vishnu wrote:
> 
> 
> On 19/11/19 6:05 PM, Dan Carpenter wrote:
>> I can't apply this because I'm not CC'd on patches 2-5.
>>
>> On Tue, Nov 19, 2019 at 05:41:16PM +0530, Ravulapati Vishnu vardhan 
>> rao wrote:
>>> +static int acp3x_power_on(void __iomem *acp3x_base)
>>> +{
>>> +    u32 val;
>>> +    u32 timeout;
>>> +
>>> +    timeout = 0;
>>> +    val = rv_readl(acp3x_base + mmACP_PGFSM_STATUS);
>>> +
>>> +    if (val == 0)
>>> +        return val;
>>> +
>>> +    if (!((val & ACP_PGFSM_STATUS_MASK) ==
>>> +                ACP_POWER_ON_IN_PROGRESS))
>>> +        rv_writel(ACP_PGFSM_CNTL_POWER_ON_MASK,
>>> +            acp3x_base + mmACP_PGFSM_CONTROL);
>>> +    while (++timeout) {
>>
>> while (++timeout < 500)
>>
> 
> If I check with timeout<500 and in next condition i have
> if(timeout >500) this never happens.
> 
> Our intention is to wait for time out and exit.
> 
>>> +        val  = rv_readl(acp3x_base + mmACP_PGFSM_STATUS);
>>                     ^^
>> Extra space character.
>>
>>
>>> +        if (!val)
>>> +            break;
>>
>> return 0;
>>
>>> +        udelay(1);
>>> +        if (timeout > 500) {
>>> +            pr_err("ACP is Not Powered ON\n");
>>> +            return -ETIMEDOUT;
>>> +        }
>>> +    }
>>> +    return 0;
>>
>> Since we combined the ++timeout and the < 500 this becomes
>> "return -ETIMEOUT;" here.
>>
>>> +}
>>> +
>>> +static int acp3x_power_off(void __iomem *acp3x_base)
>>> +{
>>> +    u32 val;
>>> +    u32 timeout, ret;
>>
>> Both ret and timeout should just be int.  Please update this throughout.
>>
>>> +
>>> +    timeout = 0;
>>
>> Move the timeout = 0 next to the loop or put it in the initializer.
>>
>>> +    rv_writel(ACP_PGFSM_CNTL_POWER_OFF_MASK,
>>> +            acp3x_base + mmACP_PGFSM_CONTROL);
>>> +    while (++timeout) {
>>
>> while (++timeout < 500) {
>>
>>> +        val  = rv_readl(acp3x_base + mmACP_PGFSM_STATUS);
>>
>> Extra space char.
>>
>>> +        if ((val & ACP_PGFSM_STATUS_MASK) == ACP_POWERED_OFF) {
>>> +            ret = 0;
>>> +            break;
>>
>> return 0;
>>
>>> +        }
>>> +        udelay(1);
>>> +        if (timeout > 500) {
>>> +            pr_err("ACP is Not Powered OFF\n");
>>> +            ret = -ETIMEDOUT;
>>> +            break;
>>> +        }
>>> +    }
>>> +    return ret;
>>> +}
>>> +
>>> +static int acp3x_reset(void __iomem *acp3x_base)
>>> +{
>>> +    u32 val, timeout;
>>> +
>>> +    rv_writel(1, acp3x_base + mmACP_SOFT_RESET);
>>> +    timeout = 0;
>>> +    while (++timeout) {
>>> +        val = rv_readl(acp3x_base + mmACP_SOFT_RESET);
>>> +        if ((val & ACP3x_SOFT_RESET__SoftResetAudDone_MASK) ||
>>> +                            timeout > 100) {
>>
>> This timeout > 100 limit was difficult to spot.  Like finding Waldo.
>>
>>> +            if (val & ACP3x_SOFT_RESET__SoftResetAudDone_MASK)
>>> +                break;
>>
>> This is a duplicate condition.
>>
>>> +            return -ENODEV;
>>> +        }
>>> +        cpu_relax();
>>> +    }
>>> +    rv_writel(0, acp3x_base + mmACP_SOFT_RESET);
>>> +    timeout = 0;
>>> +    while (++timeout) {
>>> +        val = rv_readl(acp3x_base + mmACP_SOFT_RESET);
>>> +        if (!val)
>>> +            break;
>>> +        if (timeout > 100)
>>> +            return -ENODEV;
>>> +        cpu_relax();
>>> +    }
>>> +    return 0;
>>> +}
>>> +
>>> +static int acp3x_init(void __iomem *acp3x_base)
>>> +{
>>> +    int ret;
>>> +
>>> +    /* power on */
>>> +    ret = acp3x_power_on(acp3x_base);
>>> +    if (ret) {
>>> +        pr_err("ACP3x power on failed\n");
>>> +        return ret;
>>> +    }
>>> +    /* Reset */
>>> +    ret = acp3x_reset(acp3x_base);
>>> +    if (ret) {
>>> +        pr_err("ACP3x reset failed\n");
>>> +        return ret;
>>> +    }
>>> +    return 0;
>>> +}
>>> +
>>> +static int acp3x_deinit(void __iomem *acp3x_base)
>>> +{
>>> +    int ret;
>>> +
>>> +    /* Reset */
>>> +    ret = acp3x_reset(acp3x_base);
>>> +    if (ret) {
>>> +        pr_err("ACP3x reset failed\n");
>>> +        return ret;
>>> +    }
>>> +    /* power off */
>>> +    ret = acp3x_power_off(acp3x_base);
>>> +    if (ret) {
>>> +        pr_err("ACP3x power off failed\n");
>>> +        return ret;
>>> +    }
>>> +    return 0;
>>> +}
>>> +
>>>   static int snd_acp3x_probe(struct pci_dev *pci,
>>>                  const struct pci_device_id *pci_id)
>>>   {
>>> @@ -64,6 +186,9 @@ static int snd_acp3x_probe(struct pci_dev *pci,
>>>       }
>>>       pci_set_master(pci);
>>>       pci_set_drvdata(pci, adata);
>>> +    ret = acp3x_init(adata->acp3x_base);
>>> +    if (ret)
>>> +        goto disable_msi;
>>>       val = rv_readl(adata->acp3x_base + mmACP_I2S_PIN_CONFIG);
>>>       switch (val) {
>>> @@ -73,7 +198,7 @@ static int snd_acp3x_probe(struct pci_dev *pci,
>>>                         GFP_KERNEL);
>>>           if (!adata->res) {
>>>               ret = -ENOMEM;
>>> -            goto disable_msi;
>>> +            goto de_init;
>>>           }
>>>           adata->res[0].name = "acp3x_i2s_iomem";
>>> @@ -134,12 +259,23 @@ static int snd_acp3x_probe(struct pci_dev *pci,
>>>           ret = -ENODEV;
>>>           goto disable_msi;
>>>       }
>>> +    pm_runtime_set_autosuspend_delay(&pci->dev, 5000);
>>> +    pm_runtime_use_autosuspend(&pci->dev);
>>> +    pm_runtime_set_active(&pci->dev);
>>> +    pm_runtime_put_noidle(&pci->dev);
>>> +    pm_runtime_enable(&pci->dev);
>>>       return 0;
>>>   unregister_devs:
>>>       if (val == I2S_MODE)
>>>           for (i = 0 ; i < ACP3x_DEVS ; i++)
>>>               platform_device_unregister(adata->pdev[i]);
>>> +de_init:
>>> +    ret = acp3x_deinit(adata->acp3x_base);
>>> +    if (ret)
>>> +        dev_err(&pci->dev, "ACP de-init failed\n");
>>> +    else
>>> +        dev_dbg(&pci->dev, "ACP de-initialized\n");
>>
>>
>> We can't overwrite ret (probe failed even if deinit() succeeded).  I
>> dont' know that the debug printk is useful.
>>
>> de_init:
>>     if (acp3x_deinit(adata->acp3x_base))
>>         dev_err(&pci->dev, "ACP de-init failed in probe error 
>> handling\n");
>>
>>
>>>   disable_msi:
>>>       pci_disable_msi(pci);
>>>   release_regions:
>>> @@ -150,15 +286,58 @@ static int snd_acp3x_probe(struct pci_dev *pci,
>>>       return ret;
>>>   }
>>> +static int  snd_acp3x_suspend(struct device *dev)
>>               ^^
>> Extra space char
>>
>>> +{
>>> +    int status;
>>
>> int ret;
>>
>>> +    struct acp3x_dev_data *adata;
>>> +
>>> +    adata = dev_get_drvdata(dev);
>>> +    status = acp3x_deinit(adata->acp3x_base);
>>> +    if (status)
>>> +        dev_err(dev, "ACP de-init failed\n");
>>> +    else
>>> +        dev_dbg(dev, "ACP de-initialized\n");
>>> +
>>> +    return 0;
>>> +}
>>> +
>>> +static int  snd_acp3x_resume(struct device *dev)
>>               ^^
>> Extra space
>>
>>> +{
>>> +    int status;
>>> +    struct acp3x_dev_data *adata;
>>> +
>>> +    adata = dev_get_drvdata(dev);
>>> +    status = acp3x_init(adata->acp3x_base);
>>> +    if (status) {
>>> +        dev_err(dev, "ACP init failed\n");
>>> +        return status;
>>> +    }
>>> +    return 0;
>>> +}
>>> +
>>> +static const struct dev_pm_ops acp3x_pm = {
>>> +    .runtime_suspend = snd_acp3x_suspend,
>>> +    .runtime_resume =  snd_acp3x_resume,
>>> +    .resume =       snd_acp3x_resume,
>>
>> Fix whitespace.
>>
>>> +};
>>> +
>>>   static void snd_acp3x_remove(struct pci_dev *pci)
>>>   {
>>> -    struct acp3x_dev_data *adata = pci_get_drvdata(pci);
>>
>> This was fine.  Leave it as-is.
>>

Actually I  was reported by kbuild robot tool about ISO mixed forbids of 
initialization so I did this.

>>> -    int i;
>>> +    struct acp3x_dev_data *adata;
>>> +    int i, ret;
>>> +    adata = pci_get_drvdata(pci);
>>>       if (adata->acp3x_audio_mode == ACP3x_I2S_MODE) {
>>>           for (i = 0 ; i <  ACP3x_DEVS ; i++)
>>                                  ^^
>> There is an extra space char here as well.  I guess I missed it when I
>> reviewed patch 1.
>>
>>>               platform_device_unregister(adata->pdev[i]);
>>>       }
>>> +    ret = acp3x_deinit(adata->acp3x_base);
>>> +    if (ret)
>>> +        dev_err(&pci->dev, "ACP de-init failed\n");
>>> +    else
>>> +        dev_dbg(&pci->dev, "ACP de-initialized\n");
>>
>> Put the printk in acp3x_deinit() itself and remove it from all the
>> callers.
>>
>> regards,
>> dan carpenter
>>